The close reading of participants’ responses indicated that participants used different pronouns when responding to the free recall measure. For example, whereas some participants clearly identified that the material was intended for them (e.g., “the message said that I should seek help for my depression”), other participants used impersonal or other pronouns when discussing the depression help-seeking message. It was also interesting to discover how risk and reward were discussed in response to the gain-or loss-framed messages – a theoretical framework that is based on the concept of perceived risk of behavioral outcomes. In some cases, participants discussed the positive outcomes of help-seeking despite being exposed to the loss-framed help-seeking message (i.e., the message displaying the negative outcomes of not seeking help). Finally, whereas the visual affective cues displayed in the gain-and loss message seemed to be reflected in the open-ended text responses, some participants discussed negative emotions despite being assigned to the gain-framed message.
Memory for people did not differ by message condition, Χ2 (2) = 3.90, p = .824 or depression category, Χ2 (4) = 4.62, p = .329. Whereas those with minimal depression
showed no difference in memory for people in the gain (66.7%) and loss message
(65.1%), those with mild depression and moderate to severe depression showed improved memory for people in the loss message (Memory People Mild Depression 57.1%, Memory People Moderate – Severe Depression 50%) over the gain message (Memory People Mild Depression
44.4%, Memory People Moderate – Severe Depression 33.3%)
Memory for happy and sad emotions differed by message condition, Χ2 (2) = 62.59, p < .001, but not by depression category, Χ2 (4) = 3.72, p = .445. Overall, both the minimal and mild depression groups showed better memory for happiness than sadness, whereas the moderate to severe depression group did not differ in memory for happiness and sadness (see table 17).
Memory for gain-and loss statements differed by message condition, Χ2 (4) = 60.05, p < .001, but not by depression category, Χ2 (8) = 4.52, p = .808. All depression groups reported better memory for the positive outcomes of help-seeking than the negative outcomes of help-seeking, but the difference was strongest for the minimal depression group (see table 17).
Memory for behavioral recommendation differed slightly by message condition, Χ2 (3) = 7.50, p = .058, but not by depression category, Χ2 (6) = 3.45, p = .751.All depression groups showed better memory for the recommended help-seeking behavior in the gain-message condition (see table 17).
Memory for disease information did not differ by message condition, Χ2 (1) = 0, p
= .993 or depression category, Χ2 (2) = 2.51, p = .286 (see table 17).
Table 17
Memory for Gain-and Loss-Framed Message Condition Components by Depression
Message Frame
Dependent Variable Depression Gain Frame (% within depression)
Computerized linguistic analysis results. Participants’ open-ended text responses to the free recall question, “What do you remember from the message you saw? Please
list all things (e.g., words, images, objects) that come to mind,” were analyzed with linguistic inquiry and word count software (LIWC, 2015).
Overall, an average of 6.83% of all participants’ text responses contained pronoun words (i.e. either personal or impersonal pronouns, or specific categorizations, such as
“I,” “we,” “you,” “he/she,” and “they”). 9.66% of all participants’ text responses
contained words related to affect (i.e. either general positive or negative emotion words, or specific categorizations for negative emotions, such as anxiety and sadness). Due to the nature of gain-and loss framing as focusing on positive and negative outcomes of help-seeking matched with positive and negative visual affective cues, the analysis of text responses focused on the general categorization distinguishing positive and negative emotions only.
10.06% of all participants’ text responses were drive words (i.e. words related to affiliation, achievement, power, reward, and risk). Due to the central relevance of reward and risk words in relation to gain-and loss framed health messages, analysis focused on only risk and reward words for drive.
Use of positive emotion words did not differ as a function of the interaction of message condition and depression, F(2, 139) = .33, p = .717, η2 = .01. Use of positive emotion words did also not differ by depression, F(2, 139) = .10, p = .906, η2 = .00, but by message condition, F(1, 139) = 28.83, p < .001, η2 = .17.
Similarly, use of negative emotion words did not differ as a function of the
interaction of message condition and depression, F(2, 139) = 1.19, p = .308, η2 = .02. Use of negative words did also not differ by depression, F(2, 139) = .43, p = .652, η2 = .01 and message condition, F(1, 139) = 2.74, p = .100, η2 = .02.
Findings reveal that all depression groups used positive and negative emotion words according with positive (gain) and negative (loss) framing conditions, except the mild depression group, who used slightly more negative emotion words in the gain-framed condition than negative emotion words in the loss-framed message condition (see figure 27 and 28).
Figure 27. Mean Proportion of Use of Positive Emotion Words by Depression and Message Condition.
Figure 28. Mean Proportion of Use of Negative Emotion Words by Depression and Message Condition.
Use of reward words did not differ as a function of the interaction of message condition and depression, F(2, 139) = .69, p = .503, η2 = .01. Use of reward words did also not differ by depression, F(2, 139) = .87, p = .423, η2 = .01, but by message condition, F(1, 139) = 8.08, p = .005, η2 = .06.
All depression groups used more reward words in the gain message condition, whereas the difference in use reward words in the gain message condition and loss message condition was most meaningful in the mild depression group (see figure 29).
Use of risk words, however, did differ as a function of both message condition and depression, F(2, 139) = 3.16, p = .045, η2 = .04. Use of risk words did not differ by depression, F(2, 139) = 1.67, p = .193, η2 = .02, but by message condition, F(1, 139) = 6.95, p = .009, η2 = .05.
Whereas the minimal depression groups did not differ in use of risk words in the gain-framed message condition (M = .26, SD = .92) and the loss-framed message condition (M = .24, SD = .67), the mild depression group and the moderate depression group showed interesting patterns in regard to risk word usage. For the mild depression group, use of risk words was higher in the loss-framed message condition, (M = .69, SD = 1.40) than in the gain-framed message condition, (M = .13, SD = .55). Use of risk words increased for the moderate depression to severe depression group, as well, but the difference was remarkable – whereas the moderate to severe depression group used virtually no risk words in the gain-framed message condition, (M = 00, SD = 00), risk word usage was high in the loss-framed message condition (M = 1.91, SD = 4.50) (see figure 30 and table 18).
Figure 29. Mean Proportion of Use of Reward Words by Depression and Message Condition.
Figure 30. Mean Proportion of Use of Risk Words by Depression and Message Condition.
Table 18
Mean Proportions of Word Categories in Free Recall Measure (N = 154)
Message Frame
Dependent Variable Depression Gain Frame Loss Frame Positive Emotions Minimal 5.55 (5.17) .83 (1.57)
Mild 4.88 (3.72) 1.22 (1.69)
Mod-Sev 5.32 (5.32) 1.69 (4.46) Negative Emotions Minimal 4.19 (4.37) 9.44 (15.74)
Mild 5.56 (3.81) 5.29 (2.81)
Mod-Sev 5.53 (5.47) 9.39 (7.46)
Reward Minimal 2.81 (3.16) 1.36 (2.63)
Mild 2.87 (3.54) .60 (1.15)
Mod-Sev 1.60 (1.77) .96 (1.44)
Risk Minimal .26 (.92) .24 (.67)
Mild .13 (.55) .69 (1.40)
Mod-Sev 0 (0) 1.91 (4.50)
Note: Means reflect percentages.
Discussion Stages of Cognitive Processing
Attention. Overall, the present study suggests several important implications for the design of depression help-seeking messages. Health messages in general suffer from a significant disadvantage – they must break through the vast number of competing
messages and get the viewer’s attention. Even when this objective is achieved and the health message is being attended to, the time window during which information is absorbed by the viewer is incredibly short. In the present study, participants viewed the depression help-seeking messages for an average of about thirty seconds. This duration time is likely influenced by the nature of the experimental setting – a controlled lab without distractions and the potential influence of the participant’s desire to fulfill the researcher’s expectations. We must assume that viewing duration is drastically shorter for health message viewing scenarios outside of this experimental setting.
Despite this realization, the present study also suggests several potential
opportunities for depression help-seeking messages. Whereas viewing time decreased when depression symptoms increased, the minimal depression group viewed the help-seeking messages for the longest duration of time. This is encouraging, given the objective to reach those who suffer from depressive symptoms early and to prevent worsening of symptoms. This group of students -for which depressive symptoms might just be noticeable enough to prompt awareness but not severe enough to induce resistance to help-seeking - might be an important group to target with depression help-seeking messages. Not only can a worsening of symptoms be prevented, but members of this group could also serve as important catalysts for spreading depression health information
to those in need in their social networks.
Interestingly and contrary to expectations, there were no differences in how
depression influenced attention for negative, loss-framed information and negative facial cues with one exception – the time it takes participants to fixate on message components.
It appears that those low in depression are more likely than those with higher levels of depression to quickly attend to positive visuals and information if given the opportunity.
Conversely, those with higher levels of depression seem to avoid such information in attention sequences. Interestingly, and potentially due to increased self-relevance, those with higher levels of depression appear to focus on disease information more quickly than those with lower levels of depression when negativity is salient. These patterns are suggestive of both a cognitive bias for positive information among those with lower levels of depression, as well as a cognitive bias for negative information among those with higher levels of depression.
The important difference for attention processes influenced by depression might not only be the valence of information, but also the required level of processing. For
example, mean levels indicate that those with low levels of depression focused on
positive information that allowed for quick and easy processing much more often than the high depression group – especially in regard to the headline and visual. The high
depression group, on the contrary, appeared to very quickly identify actual text-based information in the help-seeking messages. This group fixated on disease information in the gain and loss message conditions much more quickly than the low depression group.
This finding could have implications for cognitive processing mechanisms as depicted by the elaboration likelihood model (Cacioppo & Petty, 1984). It is possible that those with
high levels of depression recognized the self-relevance of the depression help-seeking message, which increased the desire for substantial and strong arguments and information (Petty & Cacioppo, 1986; Petty & Cacioppo, 1990; Petty, Haugtvedt, & Smith, 1995;
Petty, Wegener, & Fabrigar, 1997), whereas those with low depression extracted useful information by identifying the shortcuts to message processing – the headline and visual.
Independent of message framing, higher levels of depression could have also induced the more systematic and careful processing of information (Gollan et al., 2008; Petty et al., 1993).
Interpretation. Interestingly, no meaningful effects were found for interpretation of gain-and loss-framed messages for the minimal depression group. Despite viewing the depression help-seeking messages for the longest duration of time, it does not appear that message framing meaningfully influences interpretation for those with such low levels of depression. Message framing seemingly starts to make a difference starting with those who suffer from mild symptoms of depression. Realizing that messages rarely shape intentions to engage in a health behavior directly, the strong effects on intentions to seek help as shaped by message framing were surprising. Whereas the mild depression group indicated higher intentions to seek help after viewing the gain-framed message than the moderate depression group, these effects were reversed for the loss-framed message.
Here, those with moderate depression indicated higher intentions to seek help after viewing the loss-framed message than those with mild symptoms of depression – raising the means for intentions to seek help above the scale midpoint. The persuasive advantage of the gain-framed message for the mild depression group and the loss-framed message for the moderate depression group was largely confirmed for the remaining reasoned
action components.
It could be the case that perceived level of risk in regard to help-seeking outcomes plays an important role in these differences. Perceived risk refers to outcome certainty and has previously been applied in the context of gain-and loss framing. More
specifically, perceived risk has been used to distinguish health behaviors into either detection or prevention behaviors. Detection behaviors are high in perceived risk, because they typically involve high outcome uncertainty (e.g., breast cancer screening) and prevention behaviors are low in perceived risk, because they typically lead to certain outcomes (e.g., sunscreen use).
Applied to the present research, if help-seeking for depression is expected to lead to a certain outcome, such as receiving professional advice on how to improve mental health by exercising, improving sleep schedules and using stress management techniques, perceived risk would be low. If, however, a person were not sure if a visit with a health professional might lead to an unpleasant depression diagnosis, perceived risk would be high. Although not consistently confirmed in the literature and therefore recently called into question (O’Keefe & Jensen, 2007; O’Keefe & Jensen, 2009), gain-framing is believed to be more effective for prevention behaviors, which are low in perceived levels of risk, whereas loss-framing is believed to be more effective for detection behaviors, which are high in perceived levels of risk.
The present research sought to test the persuasive advantage of gain-and loss framing for help-seeking messages and those individuals who are affected by depression.
To date, no study has investigated whether depression influences the levels of perceived risk for gain-and loss framed depression help-seeking messages. The current results
support the conclusion that the gain frame is more persuasive for prevention
behaviors (i.e. help-seeking for those with low levels of depression) and that the loss frame is more persuasive for detection behaviors (i.e. help-seeking for those with high levels of depression). Furthermore, it also appears that negative information improved feelings about help-seeking and perceptions that important others would seek help themselves among those with higher levels of depression – determinants of help-seeking that were generally low and require improvement. These findings suggests that contrary to expectations, matching negatively biased cognition with negative information can indeed lead to positive outcomes, rather than additive negative effects and unintended negative consequences.
Memory. Overall, those low in depression showed slightly better memory for happy people and emotions than sad people and emotions, whereas those high in depression showed better memory for sad people and emotions. Interestingly, all
depression groups reported better memory for the positive outcomes of help-seeking than the negative outcomes of not seeking help. Whereas there were no differences in memory for disease information in the loss-framed and gain-framed message conditions among those with low depression, those with high depression reported better memory for disease information in the loss condition than in the gain condition.
These findings suggest a positivity bias for those with lower levels of depression and a negativity bias for those with higher levels of information – biases that might also influence processing style, and ultimately, information storage in memory. Improved memory for disease information among those with higher levels of depression could indicate increased message relevance for those who did indeed suffer from depression
when confronted with a message that encouraged help-seeking for depression.
Remarkably, those with higher levels of depression appeared to be highly affected by levels of perceived risk. Using their own words in direct response to gain-and loss – framed depression help-seeking messages, the moderate depression group used risk words much more often in response to the loss-framed message than in response to the gain-framed message. The mild depression group appears to be affects by perceived risk as well, but to a lesser extend. There were no differences in use of risk words in response to the gain- and loss framed messages for the minimal depression group, indicating once again that meaningful effects of message framing only seemed to emerge starting with mild symptoms of depression.
The high levels of perceived risk among those with moderate depression after viewing the loss-frame message could have several potential reasons. For one, most students in the current sample indicated that they have not previously sought help for depression. This could mean that the outcomes of help-seeking are perceived to be particularly risky for those who need help the most. Perhaps the potential of receiving a formal diagnosis, a label that carries social stigma, is especially unsettling for this moderate depression group. The uncertainty could also refer to not knowing what
actually happens during a visit with a mental health professional. Negative rumination, as one symptom of depression, could induce multiple ‘what – if scenarios’ in this regard.
For example, rumination negativity and uncertainty could range from the perceived inability to describe depression symptoms accurately, to the anticipated reaction of the health professional, to suggested treatment options that might include medication All of these questions raise levels of uncertainty and perceived risks of help-seeking.
Overall, barring careful interpretation of correlational findings, perceived risk could be the explanatory mechanisms for the positive effects of the loss-framed message on help-seeking intentions for those who are moderately depressed. Arguably, because perceived risk of help-seeking might be lower for those who only suffer from mild depression symptoms, gain frames could be more effective for this group, instead.
Returning to the prevention and detection behavior categorization (Rothman et al., 2006), the present research suggests that help-seeking for depression is likely perceived to be a prevention behavior by those who suffer form lower levels of depression and a detection behavior by those who suffer from higher levels of depression. Therefore, a gain-framed depression help-seeking message should be used for early interventions targeting those suffering from less severe depression symptoms, whereas a loss-framed depression help-seeking message should be utilized for those who urgently need help for depression.
In order to make these choices, interventions must ultimately be informed by specific knowledge regarding the target audience. Messages must be tailored and targeted in correspondence with levels of depression severity –knowledge that might aid in
constructing health messages that can induce behavior change in holistic ways. Different message strategies could work together holistically, by influencing potential opinion leaders and catalysts of stigma reduction of help-seeking for depression via gain-framed messages, or directly, by increasing help-seeking and potentially saving lives for those with high levels of depression via loss-framed messages.
Limitations
This research study is not without limitations. Most significantly affecting the patterns of results is the lack of adequate power in each of the experimental cells.
Depression categories were combined into minimal, mild, and moderate to severe depression groups in order to detect potential differences in effects and aid comparisons of unequal sample sizes. Ideally, a study sample would provide enough cases for each level of depression categories (no, minimal, mild, moderate, moderately severe, and severe) in order to conduct meaningful comparisons between levels of depression.
The issue of power most significantly affected human content analyses of participants’ open-ended responses. Here, cases for each of the coding categories as provided by groups of depression were lower than desired. Therefore, assumptions necessary for inferential statistics (i.e. logistic regression) were not supported.
Given the multiple methods employed in this study, each method also carried limitations.
For example, it is plausible that an in-person lab experiment led to certain levels of social
For example, it is plausible that an in-person lab experiment led to certain levels of social